Merchant information recommendation method, model training method, device and equipment

Abstract

The embodiment of the present application provides a merchant information recommendation method, model training method, device and equipment. The method includes: obtaining the merchant name of the user, and determining at least one target merchant name corresponding to the merchant name from the preset merchant library; wherein the preset merchant library includes at least one preset merchant name; the target merchant name is a preset merchant name associated with the merchant name; according to each target merchant name, a prompt word is generated; wherein the prompt word includes the merchant category name and context information of each target merchant name; language processing is performed on the prompt word to obtain the merchant information of the user; wherein the merchant information represents the merchant category code to be recommended; and recommending the merchant information of the user to the user. The method is used to achieve the effect of improving the accuracy of the merchant information recommended to the user.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a merchant information recommendation method, model training method, device and equipment. Background Art

[0002] When users apply for network access based on the merchant network access platform, they need to fill in merchant category code (MCC) and other merchant information; since there are many MCC categories and the classification is complex, it is necessary to recommend the MCC to which the user belongs.

[0003] In the prior art, a small pre-trained model is used to recommend 3-5 MCCs, and then a suitable MCC is manually screened out from the 3-5 MCCs.

[0004] However, in the above method, the pre-trained small model is trained based on less data, and it is difficult to capture the user's existing business characteristics, which leads to inaccurate MCC recommendations to users. Summary of the invention

[0005] The embodiments of the present application provide a merchant information recommendation method, model training method, device and equipment to achieve the effect of improving the accuracy of merchant information recommended to users.

[0006] In a first aspect, an embodiment of the present application provides a method for recommending merchant information, including:

[0007] Acquire the merchant name of the user, and determine at least one target merchant name corresponding to the merchant name from a preset merchant library; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name;

[0008] Generate prompt words according to each of the target merchant names; wherein the prompt words include merchant category names and context information of each of the target merchant names;

[0009] Performing language processing on the prompt word to obtain the merchant information of the user; wherein the merchant information represents the category code of the merchant to be recommended;

[0010] Recommending the user's business information to the user.

[0011] In a possible implementation, determining at least one target merchant name corresponding to the merchant name from a preset merchant library includes:

[0012] According to the merchant name, each of the preset merchant names is preliminarily sorted to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name;

[0013] According to the merchant name, the similar merchant names are reordered to obtain the target merchant names.

[0014] In a possible implementation, the preset merchant names are preliminarily sorted according to the merchant names to obtain a merchant name set, including:

[0015] Determining the similarity between the vector of the preset merchant name and the vector of the merchant name;

[0016] The merchant name set is determined according to each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0017] In a possible implementation, the similar merchant names are reordered according to the merchant names to obtain the target merchant names, including:

[0018] Based on the attention mechanism, the fields in the similar merchant names and the fields in the merchant name are processed to obtain association information; wherein the association information represents the degree of association between the similar merchant names and the merchant name;

[0019] According to the association information, each of the target merchant names is determined from the similar merchant names.

[0020] In a possible implementation, performing language processing on the prompt word to obtain the merchant information of the user includes:

[0021] Based on the recommendation model, the prompt word is language processed to obtain an output result; wherein the output result includes the name of the merchant category to be recommended corresponding to the merchant name;

[0022] The output result is post-processed to obtain the merchant information of the user.

[0023] In a possible implementation, post-processing the output result to obtain the merchant information of the user includes:

[0024] Parsing the output result to obtain key information; wherein the key information includes the merchant category name corresponding to the merchant name;

[0025] Code mapping is performed on the key information to obtain the merchant information of the user.

[0026] In a possible implementation, the output result is parsed to obtain key information, including:

[0027] Performing format conversion processing on the output result to obtain a converted output result;

[0028] The converted output result is subjected to information extraction processing to obtain the key information.

[0029] In a possible implementation, the method further includes:

[0030] The merchant information of the user is verified to obtain a verification result; wherein the verification result represents the accuracy of the merchant information of the user.

[0031] In a possible implementation manner, before determining at least one target merchant name corresponding to the merchant name from a preset merchant database, the method further includes:

[0032] The merchant name is preprocessed to obtain a processed merchant name.

[0033] In a second aspect, an embodiment of the present application provides a model training method for merchant information recommendation, including:

[0034] Obtain the merchant name of the user to be trained, and determine at least one target merchant name corresponding to the merchant name from a preset training set; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name;

[0035] Generate prompt words according to each of the target merchant names; wherein the prompt words include merchant category names and context information of each of the target merchant names;

[0036] According to the prompt word, the initial model is trained to obtain a recommendation model; wherein the recommendation model is used to process the merchant name of the user as described in the first aspect to obtain the merchant information of the user; the merchant information represents the merchant category code to be recommended.

[0037] In a possible implementation, determining at least one target merchant name corresponding to the merchant name from a preset training set includes:

[0038] According to the merchant name, each of the preset merchant names is preliminarily sorted to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name;

[0039] According to the merchant name, the similar merchant names are reordered to obtain the target merchant names.

[0040] In a possible implementation, the preset merchant names are preliminarily sorted according to the merchant names to obtain a merchant name set, including:

[0041] Determining the similarity between the vector of the preset merchant name and the vector of the merchant name;

[0042] The merchant name set is determined according to each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0043] In a possible implementation, the similar merchant names are reordered according to the merchant names to obtain the target merchant names, including:

[0044] Based on the attention mechanism, each of the similar merchant names and the merchant name is processed to obtain association information; wherein the association information represents the degree of association between each of the similar merchant names and the merchant name;

[0045] According to the association information, each of the target merchant names is determined from the similar merchant names.

[0046] In a possible implementation, training the initial model according to the prompt word to obtain a recommendation model includes:

[0047] Based on the initial model, the prompt word is language processed to obtain an output result; wherein the output result includes the name of the merchant category to be recommended corresponding to the merchant name;

[0048] The initial model is trained according to the merchant category name to be recommended and the actual category name corresponding to the merchant name to obtain the recommendation model; wherein the actual category name is the actual merchant category name corresponding to the merchant name of the user to be trained.

[0049] In a possible implementation manner, before determining at least one target merchant name corresponding to the merchant name from a preset training set, the method further includes:

[0050] The merchant name is preprocessed to obtain a processed merchant name.

[0051] In a third aspect, an embodiment of the present application provides a device for recommending merchant information, including:

[0052] A determination module, used to obtain a merchant name of a user, and determine at least one target merchant name corresponding to the merchant name from a preset merchant library; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name;

[0053] A generating module, used to generate prompt words according to each of the target merchant names; wherein the prompt words include the merchant category name and context information of each of the target merchant names;

[0054] A processing module, used for performing language processing on the prompt word to obtain the merchant information of the user; wherein the merchant information represents the category code of the merchant to be recommended;

[0055] The recommendation module is used to recommend the user's business information to the user.

[0056] In a possible implementation, the determination module is specifically used to: perform preliminary sorting processing on each of the preset merchant names according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; and perform re-sorting processing on each of the similar merchant names according to the merchant name to obtain each of the target merchant names.

[0057] In a possible implementation, the determination module is specifically used to: determine the similarity between the vector of the preset merchant name and the vector of the merchant name; determine the merchant name set based on each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0058] In a possible implementation, the determination module is further specifically used to: based on an attention mechanism, process the fields in each of the similar merchant names and the fields in the merchant name to obtain association information; wherein the association information represents the degree of association between each of the similar merchant names and the merchant name; and determine each of the target merchant names from each of the similar merchant names based on the association information.

[0059] In a possible implementation, the processing module is specifically used to: perform language processing on the prompt word based on a recommendation model to obtain an output result; wherein the output result includes the name of the merchant category to be recommended corresponding to the merchant name; and post-process the output result to obtain the merchant information of the user.

[0060] In a possible implementation, the processing module is specifically used to: parse the output result to obtain key information; wherein the key information includes the merchant category name corresponding to the merchant name; perform code mapping on the key information to obtain the merchant information of the user.

[0061] In a possible implementation, the processing module is specifically used to: perform format conversion processing on the output result to obtain a converted output result; and perform information extraction processing on the converted output result to obtain the key information.

[0062] In a possible implementation manner, the processing module is further specifically configured to: perform verification processing on the merchant information of the user to obtain a verification result; wherein the verification result represents the accuracy of the merchant information of the user.

[0063] In a possible implementation manner, before the device is used to determine at least one target merchant name corresponding to the merchant name from a preset merchant database, the device is further used to: pre-process the merchant name to obtain a processed merchant name.

[0064] In a fourth aspect, an embodiment of the present application provides a model training device for merchant information recommendation, including:

[0065] A determination module, used to obtain the merchant name of the user to be trained, and determine at least one target merchant name corresponding to the merchant name from a preset training set; wherein the preset merchant library includes at least one preset merchant name; the target merchant name is a preset merchant name associated with the merchant name;

[0066] A generating module, used to generate prompt words according to each of the target merchant names; wherein the prompt words include the merchant category name and context information of each of the target merchant names;

[0067] A training module is used to train the initial model according to the prompt word to obtain a recommendation model; wherein the recommendation model is used to process the merchant name of the user as described in the third aspect to obtain the merchant information of the user; the merchant information represents the merchant category code to be recommended.

[0068] In a possible implementation, the determination module is specifically used to: perform preliminary sorting processing on each of the preset merchant names according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; and perform re-sorting processing on each of the similar merchant names according to the merchant name to obtain each of the target merchant names.

[0069] In a possible implementation, the determination module is specifically used to: determine the similarity between the vector of the preset merchant name and the vector of the merchant name; determine the merchant name set based on each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0070] In a possible implementation, the determination module is further specifically used to: based on an attention mechanism, process each of the similar merchant names and the merchant name to obtain association information; wherein the association information represents the degree of association between each of the similar merchant names and the merchant name; and determine each of the target merchant names from each of the similar merchant names based on the association information.

[0071] In a possible implementation, the training module is specifically used to: perform language processing on the prompt word based on the initial model to obtain an output result; wherein the output result includes the merchant category name to be recommended corresponding to the merchant name; perform training processing on the initial model according to the merchant category name to be recommended corresponding to the merchant name and the actual category name to obtain the recommendation model; wherein the actual category name is the actual merchant category name corresponding to the merchant name of the user to be trained.

[0072] In a possible implementation, before the device is used to determine at least one target merchant name corresponding to the merchant name from a preset training set, the device is further used to: pre-process the merchant name to obtain a processed merchant name.

[0073] In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0074] The memory stores computer-executable instructions;

[0075] The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method of the first aspect and / or the second aspect as described above.

[0076] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-readable storage medium is stored computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method of the first aspect and / or the second aspect as described above.

[0077] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method of the first aspect and / or the second aspect as described above.

[0078] The merchant information recommendation method, model training method, device and equipment provided in the embodiments of the present application retrieve preset merchant names associated with the user's merchant name through merchant database retrieval, and input each associated preset merchant name as part of the prompt word into the large model to obtain the merchant category code to be recommended, and recommend it to the user, thereby achieving the effect of improving the accuracy of the merchant information recommended to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0080] Figure 1 A schematic diagram of an application scenario provided for this application;

[0081] Figure 2 A flowchart of a method for recommending merchant information provided in an embodiment of the present application;

[0082] Figure 3 A flowchart of another method for recommending merchant information provided in an embodiment of the present application;

[0083] Figure 4 A large model MCC recommendation structure diagram based on retrieval enhancement and memory provided in an embodiment of the present application;

[0084] Figure 5 A flowchart of a model training method for merchant information recommendation provided in an embodiment of the present application;

[0085] Figure 6 A flowchart of another model training method for merchant information recommendation provided in an embodiment of the present application;

[0086] Figure 7 A schematic diagram of the structure of a device for recommending merchant information provided in an embodiment of the present application;

[0087] Figure 8 A schematic diagram of the structure of a model training device for merchant information recommendation provided in an embodiment of the present application;

[0088] Fig. 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0089] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0090] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0091] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0092] In addition, this application involves conducting big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and using artificial intelligence technology to make automated decisions, and making technical solutions that have a significant impact on personal rights and interests based on the results of automated decisions. The application provides users with corresponding operation entrances for them to choose to agree or reject the results of automated decisions; if the user chooses to reject, the expert decision-making process will be entered.

[0093] It should be noted that the present application can be used in the field of artificial intelligence, and can also be used in any field other than artificial intelligence, and the application field of the present application is not limited.

[0094] Figure 1 A schematic diagram of an application scenario provided for this application, such as Figure 1As shown, the specific application scenario of the present application includes an electronic device 101 and a user 102. When the user 102 applies for network access on the merchant network access platform based on the electronic device 101, the user 102 needs to fill in the merchant information such as the MCC; since there are many categories of MCC and the classification is complex, it is necessary to recommend the MCC to the user based on the electronic device 101.

[0095] Based on the above scenarios, it can be seen that in the existing technology, secondary training is performed through pre-training small models such as Bidirectional Encoder Representations from Transformers (BERT), and the recommended MCC is mostly 3-5, which requires manual screening from 3-5 options.

[0096] However, in the above method, the pre-trained small model is trained based on less data, and it is difficult to capture the characteristics of the user's existing merchant names and business scenarios; and due to data and parameter limitations, the small model performs poorly when generalized to new or uncommon merchant types, and may not be able to accurately identify and recommend MCCs suitable for specific merchants; thus, the accuracy and reliability of the recommendation results are reduced.

[0097] The merchant information recommendation method, model training method, device and equipment provided in the present application retrieve preset merchant names associated with the user's merchant name by searching through the merchant database, and input each associated preset merchant name as part of the prompt word into the large model to obtain the merchant category code to be recommended, and recommend it to the user, thereby solving the technical problem of inaccurate MCC recommended to the user based on the pre-trained small model.

[0098] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0099] Figure 2 A flowchart of a method for recommending merchant information provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the method includes:

[0100] 201. Obtain a merchant name of a user, and determine at least one target merchant name corresponding to the merchant name from a preset merchant library; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name.

[0101] Exemplarily, the execution subject of this embodiment may be an electronic device. For ease of description, the electronic device is hereinafter referred to as the device. When the device is used to recommend merchant information, some manually screened and correctly labeled high-quality merchant business names, i.e., preset merchant names, such as some common merchant names, are first obtained, and merchant ratios of different MCC names are selected according to the merchant category names in actual business, i.e., the number of merchants with MCC names, i.e., each MCC name corresponds to a certain number of preset merchant names, for example, the number of preset merchant names in each MCC name is not less than 40, and all MCC names and corresponding preset merchant names are stored in the preset merchant library. When the user inputs the merchant name to be recommended to the device through an input operation, the device obtains the merchant name and retrieves the preset merchant library. Based on the Retrieval-Augmented Generation (RAG) technology, the device performs merchant data retrieval processing on the preset merchant database according to the merchant name entered by the user, and finds multiple preset merchant names associated with the merchant name, namely the target merchant name, and also obtains the merchant category name to which each target merchant name belongs. For example, for a new coffee shop, similar merchant names such as "A Coffee" and "B Coffee" are retrieved through RAG, and the corresponding MCC names such as "Dining Places and Restaurants" are searched at the same time to narrow the recommendation range.

[0102] 202. Generate prompt words according to each target merchant name; wherein the prompt words include merchant category names and context information of each target merchant name.

[0103] Exemplarily, the device organizes the retrieved multiple associated target merchant names and the merchant category name of each target merchant name into memory content, and concatenates the memory content with the context generated based on the user's needs to obtain a prompt word.

[0104] For example, the device retrieves the top 10 merchant names and their corresponding MCC names, splices them into the context as memory content, and uses the MCC names corresponding to the top 10 merchants as label options for the current merchant name to provide rich and diverse data input. At the same time, in order to ensure that a specific example and label option is not preferred during subsequent processing, the order of these label options and context can be randomly disrupted, and the resulting prompt words are as follows: "If you are a merchant review specialist, please label the merchant based on the merchant information. Label options: 1, MCC name: MCC definition; 2, MCC name: MCC definition; ... Example: Merchant name: XX merchant, output: dining places and restaurants; Merchant name: XX merchant, output: department store; ...top-k... merchants; Next I will give you a new merchant name, please output a label for it, the label must be selected from the given label options, the result must be returned strictly in the format of the example, and no other content should be returned. Merchant name: {business name of the merchant to be recommended}, output: XX."

[0105] 203. Perform language processing on the prompt word to obtain the user's merchant information; wherein the merchant information represents the category code of the merchant to be recommended.

[0106] Exemplarily, the device can perform language recognition processing on the obtained prompt words based on artificial intelligence technology, such as deep learning models, natural language processing models and other large models, and output the merchant category code MCC to be recommended to the user, that is, the merchant information.

[0107] 204. Recommend the user's merchant information to the user.

[0108] Exemplarily, after obtaining the user's merchant information, the device can recommend the user's merchant information to the user through interface display, voice output, etc., or send the user's merchant information to the user device currently held or used by the user through remote communication, so as to recommend the MCC corresponding to the user's merchant name to the user.

[0109] In this embodiment, a method for recommending merchant information is provided. By searching the merchant database, preset merchant names associated with the user's merchant name are retrieved, and each associated preset merchant name is input into a large model as a part of the prompt word to obtain the merchant category code to be recommended, and the merchant category code is recommended to the user, so that sufficient information is obtained in the processing process to make an accurate MCC selection, which can improve the accuracy of recommending MCC to the user.

[0110] Figure 3 A flowchart of another method for recommending merchant information provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the method includes:

[0111] 301. Get the user's merchant name.

[0112] Exemplarily, this step may refer to step 201 and will not be described in detail here.

[0113] 302. Pre-process the merchant name to obtain a processed merchant name.

[0114] For example, Figure 4 A large model MCC recommendation structure diagram based on retrieval enhancement and memory is provided in the embodiment of the present application, such as Figure 4 As shown in the figure, the process mainly consists of input preprocessing, search, merchant library, large model call and post-processing modules. The device first preprocesses the merchant name input by the user, including sensitive word filtering and stop word filtering, and obtains the processed merchant name. Among them, sensitive word filtering aims to identify and remove non-compliant words in the name. Through this process, the legality and compliance of the merchant name can be ensured; stop word filtering mainly removes some common but meaningless words in the merchant name, especially geographical location and corporate form.

[0115] For example, the device uses a sensitive word list for filtering, which includes non-compliant words; at the same time, it removes words such as "province, city, district, county, town, street, village" and "limited company, joint-stock company, group, company, enterprise, studio", etc. By filtering with stop words, the business names can be simplified, thereby improving the recall rate of the search algorithm.

[0116] 303. Preliminary sorting is performed on each preset merchant name according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name.

[0117] For example, in combination Figure 4 In order to match the associated merchant according to the input merchant name, the device selects multiple correctly labeled preset merchant names and corresponding MCC names, and stores them in the preset merchant library. Based on the preliminary sorting method in the retrieval enhancement technology, the similarity between each preset merchant name and the merchant name can be obtained, and each preset merchant name is preliminarily sorted according to the similarity, and then multiple preset merchant names with high similarity are selected from them, which are similar merchant names. Then, preset merchant names with high relevance can be preliminarily screened out.

[0118] In one example, step 303 includes the following steps:

[0119] The first step of step 303 is to determine the similarity between the vector of the preset merchant name and the vector of the merchant name.

[0120] The second step of step 303 is to determine a merchant name set according to each similarity; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant name in the preset merchant library except the merchant name set.

[0121] Exemplarily, when constructing a preset merchant database, the device performs vectorization processing on each preset merchant name based on vectorization technology, generates a vector corresponding to each preset merchant name, and stores it in the vector database part of the merchant database. The device also performs vectorization processing on the current merchant name to obtain the vector corresponding to the merchant name, and determines the degree of similarity between the merchant name and each preset merchant name by calculating the similarity between the vector corresponding to the merchant name and the vector corresponding to each preset merchant name. The device then selects a merchant name set based on the size of each similarity, and the similarity of all similar merchant names in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0122] For example, the similarity between the vector of each preset merchant name and the vector of each merchant name may be calculated by using similarity calculation methods such as cosine similarity and Euclidean distance, and the top-30 similar merchant names may be selected.

[0123] 304. Reorder similar merchant names according to the merchant names to obtain target merchant names.

[0124] For example, in combination Figure 4 Based on the reranking technology in the search enhancement technology, the device reranks the similar business names corresponding to the business name input by the user, such as ranking the correlation between the business name and each similar business name based on the rerank model, and further selects the preset business name associated with the business name from each similar business name according to the correlation, that is, obtains multiple target business names. Then, the relevant business names are retrieved from the database through preliminary sorting; then, the preliminary results are further optimized through reranking to provide more accurate search results.

[0125] In one example, step 304 includes the following steps:

[0126] The first step of step 304 is to process the fields in each similar merchant name and the fields in the merchant name based on the attention mechanism to obtain association information; wherein the association information represents the degree of association between each similar merchant name and the merchant name.

[0127] The second step of step 304 is to determine each target merchant name from each similar merchant name according to the association information.

[0128] Exemplarily, for each similar merchant name and merchant name, each of which contains multiple fields, such as the merchant name "coffee shop" contains multiple fields such as "coffee", "coffee" and "shop", the device can call a rerank model based on the attention mechanism, and based on the attention mechanism, perform attention processing on the fields in each similar merchant name and the fields in the merchant name to obtain the degree of association between each similar merchant name and the merchant name; for example, for each similar merchant name, calculate the attention weight coefficient between each field in the similar merchant name and each field in the merchant name, and then perform normalization processing to obtain the degree of association between the similar merchant name and the merchant name, and then obtain the degree of association between all similar merchant names and the merchant name. Based on the rerank model, according to the degree of correlation between each similar merchant name and the merchant name, a preset number of similar merchant names with a high degree of correlation are selected from all similar merchant names as the final target merchant name; such as preliminary sorting, selecting the top-30 similar merchant names, and selecting the rerank model for reranking can better capture similar information, and finally selecting the top-10 similar merchant names as the final matching result.

[0129] 305. Generate prompt words according to the names of the target merchants; wherein the prompt words include the merchant category name and context information of the names of the target merchants.

[0130] Exemplarily, this step may refer to step 202 and will not be described in detail here.

[0131] 306. Based on the recommendation model, perform language processing on the prompt word to obtain an output result; wherein the output result includes the merchant name corresponding to the merchant category name to be recommended.

[0132] For example, in combination Figure 4 The device calls a preset recommendation model, such as a natural language processing large model, performs language recognition processing on the obtained prompt word, and outputs the merchant category name to be recommended corresponding to the current merchant name from the MCC name to which each target merchant name in the prompt word belongs.

[0133] 307. Post-process the output result to obtain the user's merchant information.

[0134] For example, in combination Figure 4 The device performs post-processing on the output result of the recommendation model, such as some business logic processing, and obtains the corresponding MCC according to the merchant category name in the output result, and thus obtains the merchant information of the user.

[0135] In one example, step 307 includes the following steps:

[0136] The first step of step 307 is to parse the output result to obtain key information; wherein the key information includes the merchant category name corresponding to the merchant name.

[0137] The second step of step 307 is to perform code mapping processing on the key information to obtain the user's merchant information.

[0138] For example, in combination Figure 4 The device performs content parsing on the output results to obtain key information, including the merchant category name corresponding to the merchant name in the output results, and obtains the MCC corresponding to the key information through the preset mapping relationship between each MCC name and each MCC, so as to ensure that the results recommended by the model can be correctly applied to the business process

[0139] In one example, the first step of step 307 includes:

[0140] Step 1: Perform format conversion on the output result to obtain the converted output result.

[0141] Step 2: Perform information extraction on the converted output results to obtain key information.

[0142] Specifically, the device performs format conversion processing on the output result according to the model output format, such as JSON, through output format parsing, and converts the output result into a processable data structure to obtain the converted output result. Based on the information extraction technology, the device performs information extraction processing on the converted output result to obtain key information, such as the recommended MCC name and other business-related information.

[0143] In one example, step 307 further includes: verifying the merchant information of the user to obtain a verification result; wherein the verification result represents the accuracy of the merchant information of the user.

[0144] For example, the post-processing stage may also involve other business logic processing, depending on the specific application scenario and business requirements. For example, the device needs to further verify the accuracy of the MCC code in the user's merchant information, or adjust the results according to business rules to ensure the practicality and reliability of the recommendation results.

[0145] 308. Recommend the user's merchant information to the user.

[0146] Exemplarily, this step may refer to step 204 and will not be described in detail here.

[0147] In this embodiment, based on the above embodiment, on the one hand, the merchant names retrieved by RAG and their corresponding correct MCC codes can be introduced as prompt words as historical data and memory content, and integrated into the large model through the prompt words to improve the task capability of the large model, reduce the deviation of the model's understanding of the MCC definition, and better obtain reliable reasoning and output; on the other hand, through preprocessing, preliminary sorting, reordering, post-processing and other processes, it can be ensured that the recommendation model can obtain sufficient information during the processing process, thereby recommending accurate MCC.

[0148] Figure 5 A flow chart of a model training method for merchant information recommendation provided in an embodiment of the present application is shown as follows: Figure 5 As shown, the method includes:

[0149] 401. Obtain a merchant name of a user to be trained, and determine at least one target merchant name corresponding to the merchant name from a preset training set; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name.

[0150] Exemplarily, the execution subject of this embodiment may be the device in the above-mentioned embodiment. When the device is used to recommend merchant information, it may first refer to the steps for building a preset merchant library, that is, first obtain some manually screened and correctly labeled high-quality merchant business names, that is, preset merchant names, and store all MCC names and corresponding preset merchant names in a preset training set. The device obtains the merchant name of the user to be trained and retrieves the preset training set. Based on the retrieval enhancement technology, the device performs merchant data retrieval processing on the preset training set according to the merchant name input by the user to be trained, and finds multiple preset merchant names associated with the merchant name, that is, the target merchant name. At the same time, the merchant category name to which each target merchant name belongs is also obtained. It is worth adding that the establishment of the preset training set can be carried out simultaneously with the establishment of the preset merchant library, that is, a large number of manually screened and correctly labeled preset merchant names and the corresponding preset merchant names are stored separately to obtain the preset merchant library, preset training set, and preset test set; among them, in order to prevent the overfitting problem in the search and subsequent training, we split and deduplicate the data in the preset merchant library, the data in the preset training set, and the data in the preset test set to ensure that the three types of data do not overlap. It is worth adding that in order to standardize the output format and further improve the model's understanding of the MCC definition, fine-tuning is performed by using tens of thousands of annotated data, that is, the data in the preset training set includes open source general data and business scenario precision data sets. General data uses open source data sets and is mixed with precision data in a certain ratio. Business scenario precision data is mainly the answer data of merchant names and their corresponding MCC classifications. Among them, the merchant categories involved cover various MCC categories, and the data sets are evenly distributed. The corresponding answer data includes merchant categories, etc., and are all formatted as json, so that the model can standardize the output format while understanding the business judgment logic, which is convenient for later result verification and analysis.

[0151] 402. Generate prompt words according to each target merchant name; wherein the prompt words include merchant category names and context information of each target merchant name.

[0152] Exemplarily, the device organizes the retrieved multiple associated target merchant names and the merchant category name of each target merchant name into memory content, and concatenates the memory content with the context generated based on the user's needs to obtain a prompt word.

[0153] 403. According to the prompt word, the initial model is trained to obtain a recommendation model; wherein the recommendation model is used to process the user's merchant name in the merchant information recommendation method to obtain the user's merchant information; the merchant information represents the merchant category code to be recommended.

[0154] Exemplarily, the device can call initial models such as deep learning models and natural language processing models based on artificial intelligence technology, and train the initial model according to the obtained prompt words to obtain a recommendation model. The recommendation model can process the merchant name entered by the user in the actual application scenario to obtain the merchant category code MCC to be recommended by the user, that is, the merchant information, and recommend it to the user.

[0155] In this embodiment, a model training method for merchant information recommendation is provided. Preset merchant names associated with the merchant name of the user to be trained are retrieved by means of training set retrieval, and each associated preset merchant name is input into the initial model as a part of the prompt word for training processing to obtain a recommendation model with better performance. When the recommendation model is used to recommend merchant category codes to users, the accuracy of the merchant information recommended to users can be improved.

[0156] Figure 6 A flow chart of another model training method for merchant information recommendation provided in an embodiment of the present application is shown as follows: Figure 6 As shown, the method includes:

[0157] 501. Obtain the merchant name of the user to be trained.

[0158] Exemplarily, this step may refer to step 401 and will not be described in detail here.

[0159] 502. Pre-process the merchant name to obtain a processed merchant name.

[0160] For example, the device first pre-processes the merchant name input by the training user, including sensitive word filtering and stop word filtering, to obtain the processed merchant name. Among them, sensitive word filtering aims to identify and remove non-compliant words in the name. Through this process, the legality and compliance of the merchant name can be ensured; stop word filtering mainly removes some common but meaningless words in the merchant name, especially geographical location and corporate form.

[0161] 503. Preliminary sorting is performed on each preset merchant name according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name.

[0162] For example, in order to match the associated merchant according to the input merchant name, the device selects multiple correctly labeled preset merchant names and corresponding MCC names, and stores them in the preset merchant library. Based on the preliminary sorting method in the search enhancement technology, the similarity between each preset merchant name and the merchant name can be obtained, and each preset merchant name is preliminarily sorted according to the similarity, and then multiple preset merchant names with high similarity are selected from them, that is, similar merchant names. Then, the preset merchant names with high relevance to the merchant name of the user to be trained can be preliminarily screened out.

[0163] In one example, step 503 includes the following steps:

[0164] The first step of step 503 is to determine the similarity between the vector of the preset merchant name and the vector of the merchant name.

[0165] The second step of step 503 is to determine a merchant name set according to each similarity; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant name in the preset merchant library except the merchant name set.

[0166] Exemplarily, when constructing a preset training set, the device performs vectorization processing on each preset merchant name based on vectorization technology, generates a vector corresponding to each preset merchant name, and stores it in the vector database part of the merchant database. The device also performs vectorization processing on the merchant name of the user to be trained, obtains the vector corresponding to the merchant name, and determines the degree of similarity between the merchant name and each preset merchant name by calculating the similarity between the vector corresponding to the merchant name and the vector corresponding to each preset merchant name. The device then selects a merchant name set based on the size of each similarity, and the similarity of all similar merchant names in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

[0167] 504. Reorder similar merchant names according to the merchant names to obtain target merchant names.

[0168] Exemplarily, the device re-ranks the similar business names corresponding to the business name input by the user to be trained based on the re-ranking technology in the retrieval enhancement technology, such as ranking the correlation between the business name and each similar business name based on the re-ranking model, and further selecting the preset business name associated with the business name from each similar business name according to the correlation, that is, obtaining multiple target business names. Then, the relevant business names are retrieved from the database through preliminary sorting; then, the preliminary results are further optimized through re-ranking to provide more accurate search results.

[0169] In one example, step 504 includes the following steps:

[0170] The first step of step 504 is to process each similar merchant name and merchant name based on the attention mechanism to obtain associated information; wherein the associated information represents the degree of association between each similar merchant name and merchant name.

[0171] The second step of step 504 is to determine each target merchant name from each similar merchant name according to the association information.

[0172] Exemplarily, for each similar business name and the business name of the user to be trained, each of which contains multiple fields, such as the business name "coffee shop" contains multiple fields such as "coffee", "coffee" and "shop", the device can call a rerank model based on the attention mechanism, and based on the attention mechanism, perform attention processing on the fields in each similar business name and the fields in the business name to obtain the degree of association between each similar business name and the business name; for example, for each similar business name, calculate the attention weight coefficient between each field in the similar business name and each field in the business name, and then perform normalization processing to obtain the degree of association between the similar business name and the business name, and then obtain the degree of association between all similar business names and the business name. Based on the rerank model, according to the degree of correlation between each similar merchant name and the merchant name, a preset number of similar merchant names with a high degree of correlation are selected from all similar merchant names as the final target merchant name; such as preliminary sorting, selecting the top-30 similar merchant names, and selecting the rerank model for reranking can better capture similar information, and finally selecting the top-10 similar merchant names as the final matching result.

[0173] 505. Generate prompt words according to the names of the target merchants; wherein the prompt words include the merchant category name and context information of the names of the target merchants.

[0174] Exemplarily, this step may refer to step 402 and will not be described in detail here.

[0175] 506. Based on the initial model, perform language processing on the prompt word to obtain an output result; wherein the output result includes the name of the merchant category corresponding to the merchant name to be recommended.

[0176] Exemplarily, the device calls a preset initial model, such as a natural language processing large model, performs language recognition processing on the obtained prompt word, and outputs the merchant category name to be recommended corresponding to the merchant name to be trained for the user from the MCC name to which each target merchant name in the prompt word belongs.

[0177] 507. According to the merchant category name to be recommended and the actual category name corresponding to the merchant name, the initial model is trained to obtain a recommendation model; wherein the actual category name is the actual merchant category name corresponding to the merchant name of the user to be trained.

[0178] Exemplarily, the device compares the merchant name of the user to be trained with the merchant category name to be recommended and the actual merchant category name corresponding to the merchant name of the user to be trained, that is, the actual category name, and performs training processing on the initial model. For example, the corresponding loss function can be calculated to adjust the parameters of the initial model, and then obtain the recommendation model for recommendation processing of merchant information.

[0179] For example, when testing the model using the test set, the recommendation accuracy of the following methods was compared, as shown in Table 1 below.

[0180] Table 1 Comparison of recommendation accuracy

[0181]

[0182] Among them, by not using RAG technology, the traditional prompt word method is directly used for MCC recommendation. The prompt word used is "##Goal: You are a merchant audit specialist. Please label the input merchant business name from the label options. ##Requirements: 1. The label must be selected from the given label options; 2. Please output the label directly, and do not output anything else. ##Label options: 1, Label 1: MCC definition; 2, Label 2: MCC definition;... Merchant name: {Merchant business name to be recommended}, output:." Through the K-NearestNeighbor (KNN) classification algorithm, the MCC of the Top-1 similar merchant is directly used as the MCC of the current merchant. By using the proposed RAG technology, the range of MCC options is narrowed by retrieving similar merchants. Through the method of fine-tuning the initial model on the RAG strategy of this application, the initial model is fine-tuned on the basis of using the RAG strategy to improve the recommendation accuracy. The specific fine-tuning parameters are shown in Table 2 below.

[0183] Table 2 Fine-tuning parameters

[0184]

[0185] Furthermore, using the traditional prompt word method, the output accuracy of the model is only 10% because the prompt words are mixed with too much contextual information defined by MCC. When the RAG strategy proposed in this paper is adopted, its accuracy can reach 93%. The accuracy after fine-tuning can reach 97%. In addition, the difference between the accuracy of the test set and the accuracy of the validation set is less than 0.5%, indicating that the fine-tuning parameters used do not cause the model to overfit.

[0186] In this embodiment, on the basis of the above embodiments, on the one hand, since the basic large model has strong general capabilities, it still has strong reasoning capabilities when encountering new or uncommon merchant types, especially when providing reference examples as memory modules in a few-shot manner, the large model has better generalization capabilities than the small model; on the other hand, when the business logic changes, the base model of the large model does not need to be changed. Through retrieval enhancement, the new business logic and MCC definition can meet the iteration requirements only by changing the prompt words, without the need to start model training from scratch.

[0187] Figure 7 A schematic diagram of a device for recommending merchant information provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the device comprises:

[0188] The determination module 601 is used to obtain the merchant name of the user and determine at least one target merchant name corresponding to the merchant name from the preset merchant library; wherein the preset merchant library includes at least one preset merchant name; the target merchant name is a preset merchant name associated with the merchant name;

[0189] A generating module 602 is used to generate prompt words according to each target merchant name; wherein the prompt words include the merchant category name and context information of each target merchant name;

[0190] The processing module 603 is used to perform language processing on the prompt word to obtain the user's merchant information; wherein the merchant information represents the category code of the merchant to be recommended;

[0191] The recommendation module 604 is used to recommend the user's business information to the user.

[0192] In a possible implementation, the determination module 601 is specifically used to: perform preliminary sorting processing on each preset merchant name according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; according to the merchant name, re-sort the similar merchant names to obtain each target merchant name.

[0193] In a possible implementation, the determination module 601 is specifically used to: determine the similarity between the vector of the preset merchant name and the vector of the merchant name; determine the merchant name set based on each similarity; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except the merchant name set.

[0194] In a possible implementation, the determination module 601 is further specifically used to: based on the attention mechanism, process the fields in each similar merchant name and the fields in the merchant name to obtain association information; wherein the association information represents the degree of association between each similar merchant name and the merchant name; and determine each target merchant name from each similar merchant name based on the association information.

[0195] In a possible implementation, the processing module 603 is specifically used to: perform language processing on the prompt word based on the recommendation model to obtain an output result; wherein the output result includes the merchant name corresponding to the merchant category name to be recommended; and post-process the output result to obtain the user's merchant information.

[0196] In a possible implementation, the processing module 603 is specifically used to: parse the output result to obtain key information; wherein the key information includes the merchant category name corresponding to the merchant name; perform code mapping on the key information to obtain the user's merchant information.

[0197] In a possible implementation, the processing module 603 is specifically used to: perform format conversion processing on the output result to obtain a converted output result; and perform information extraction processing on the converted output result to obtain key information.

[0198] In a possible implementation, the processing module 603 is further specifically used to: verify the merchant information of the user to obtain a verification result; wherein the verification result represents the accuracy of the merchant information of the user.

[0199] In a possible implementation, before the device is used to determine at least one target merchant name corresponding to a merchant name from a preset merchant library, the device is also used to: pre-process the merchant name to obtain a processed merchant name.

[0200] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0201] Figure 8 A schematic diagram of a model training device for merchant information recommendation provided in an embodiment of the present application is shown in FIG. Figure 8 As shown, the device comprises:

[0202] The determination module 701 is used to obtain the merchant name of the user to be trained, and determine at least one target merchant name corresponding to the merchant name from the preset training set; wherein the preset merchant library includes at least one preset merchant name; the target merchant name is a preset merchant name associated with the merchant name;

[0203] A generating module 702 is used to generate prompt words according to each target merchant name; wherein the prompt words include the merchant category name and context information of each target merchant name;

[0204] The training module 703 is used to train the initial model according to the prompt word to obtain the recommendation model; wherein the recommendation model is used to process the merchant name of the user of the merchant information recommendation device to obtain the merchant information of the user; the merchant information represents the merchant category code to be recommended.

[0205] In a possible implementation, the determination module 701 is specifically used to: perform preliminary sorting processing on each preset merchant name according to the merchant name to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; according to the merchant name, re-sort the similar merchant names to obtain each target merchant name.

[0206] In a possible implementation, the determination module 701 is specifically used to: determine the similarity between the vector of the preset merchant name and the vector of the merchant name; determine the merchant name set based on each similarity; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant name in the preset merchant library except the merchant name set.

[0207] In a possible implementation, the determination module 701 is further specifically used to: based on the attention mechanism, process each similar merchant name and merchant name to obtain associated information; wherein the associated information represents the degree of association between each similar merchant name and merchant name; and determine each target merchant name from each similar merchant name based on the associated information.

[0208] In a possible implementation, the training module 703 is specifically used to: perform language processing on the prompt word based on the initial model to obtain an output result; wherein the output result includes the merchant category name to be recommended corresponding to the merchant name; perform training processing on the initial model according to the merchant category name to be recommended corresponding to the merchant name and the actual category name to obtain a recommendation model; wherein the actual category name is the actual merchant category name corresponding to the merchant name of the user to be trained.

[0209] In a possible implementation, before the device is used to determine at least one target merchant name corresponding to a merchant name from a preset training set, the device is also used to: pre-process the merchant name to obtain a processed merchant name.

[0210] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0211] Fig. 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown in FIG. Fig. 9 As shown, the electronic device includes: a memory 801 and a processor 802; the memory 801 is a memory for storing instructions executable by the processor 802.

[0212] The processor 802 is configured to execute the method provided in the above embodiment.

[0213] The electronic device 800 further includes a receiver 803 and a transmitter 804. The receiver 803 is used to receive instructions and data sent by other devices, and the transmitter 804 is used to send instructions and data to external devices.

[0214] The specific implementation process of the processor can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0215] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0216] An embodiment of the present application also provides a chip for executing instructions, which is used to execute the technical solution of the processing method in the above embodiment.

[0217] The embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed on a computer, the computer executes the technical solution of the processing method of the above embodiment.

[0218] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0219] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in a dedicated integrated circuit. Of course, the processor and the readable storage medium can also exist in a device as discrete components.

[0220] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, the technical solution of the processing method in the above embodiment can be implemented.

[0221] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0222] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0223] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0224] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for recommending merchant information, characterized in that: include: Acquire the merchant name of the user, and determine at least one target merchant name corresponding to the merchant name from a preset merchant library; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name; Generate prompt words according to each of the target merchant names; wherein the prompt words include merchant category names and context information of each of the target merchant names; Performing language processing on the prompt word to obtain the merchant information of the user; wherein the merchant information represents the category code of the merchant to be recommended; Recommending the user's business information to the user.

2. The method according to claim 1, characterized in that Determining at least one target merchant name corresponding to the merchant name from a preset merchant library includes: According to the merchant name, each of the preset merchant names is preliminarily sorted to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; According to the merchant name, the similar merchant names are reordered to obtain the target merchant names.

3. The method according to claim 2, characterized in that According to the merchant names, each of the preset merchant names is preliminarily sorted to obtain a merchant name set, including: Determining the similarity between the vector of the preset merchant name and the vector of the merchant name; The merchant name set is determined according to each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

4. The method according to claim 2, characterized in that: According to the merchant name, the similar merchant names are reordered to obtain the target merchant names, including: Based on the attention mechanism, the fields in the similar merchant names and the fields in the merchant name are processed to obtain association information; wherein the association information represents the degree of association between the similar merchant names and the merchant name; According to the association information, each of the target merchant names is determined from the similar merchant names.

5. The method according to claim 1, characterized in that: Performing language processing on the prompt word to obtain the merchant information of the user includes: Based on the recommendation model, the prompt word is language processed to obtain an output result; wherein the output result includes the name of the merchant category to be recommended corresponding to the merchant name; The output result is post-processed to obtain the merchant information of the user.

6. The method according to claim 5, characterized in that Post-processing the output result to obtain the merchant information of the user includes: Parsing the output result to obtain key information; wherein the key information includes the merchant category name corresponding to the merchant name; Code mapping is performed on the key information to obtain the merchant information of the user.

7. The method according to claim 6, characterized in that The output result is parsed to obtain key information, including: Performing format conversion processing on the output result to obtain a converted output result; The converted output result is subjected to information extraction processing to obtain the key information.

8. The method according to claim 5, characterized in that The method further comprises: The merchant information of the user is verified to obtain a verification result; wherein the verification result represents the accuracy of the merchant information of the user.

9. The method according to any one of claims 1 to 8, characterized in that Before determining at least one target merchant name corresponding to the merchant name from the preset merchant library, the method further includes: The merchant name is preprocessed to obtain a processed merchant name.

10. A model training method for merchant information recommendation, characterized in that: include: Obtain the merchant name of the user to be trained, and determine at least one target merchant name corresponding to the merchant name from a preset training set; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name; Generate prompt words according to each of the target merchant names; wherein the prompt words include merchant category names and context information of each of the target merchant names; According to the prompt word, the initial model is trained to obtain a recommendation model; wherein the recommendation model is used to process the merchant name of the user as described in any one of claims 1-9 to obtain the merchant information of the user; the merchant information represents the merchant category code to be recommended.

11. The method according to claim 10, characterized in that Determining at least one target merchant name corresponding to the merchant name from a preset training set includes: According to the merchant name, each of the preset merchant names is preliminarily sorted to obtain a merchant name set; wherein the merchant name set includes at least one similar merchant name; the similar merchant name is a preset merchant name similar to the merchant name; According to the merchant name, the similar merchant names are reordered to obtain the target merchant names.

12. The method according to claim 11, characterized in that According to the merchant names, each of the preset merchant names is preliminarily sorted to obtain a merchant name set, including: Determining the similarity between the vector of the preset merchant name and the vector of the merchant name; The merchant name set is determined according to each of the similarities; wherein the similarity corresponding to each similar merchant name in the merchant name set is greater than the similarity corresponding to the preset merchant names in the preset merchant library except for the merchant name set.

13. The method according to claim 11, characterized in that According to the merchant name, the similar merchant names are reordered to obtain the target merchant names, including: Based on the attention mechanism, each of the similar merchant names and the merchant name is processed to obtain association information; wherein the association information represents the degree of association between each of the similar merchant names and the merchant name; According to the association information, each of the target merchant names is determined from the similar merchant names.

14. The method according to claim 10, characterized in that According to the prompt words, the initial model is trained to obtain a recommendation model, including: Based on the initial model, the prompt word is language processed to obtain an output result; wherein the output result includes the name of the merchant category to be recommended corresponding to the merchant name; The initial model is trained according to the merchant category name to be recommended and the actual category name corresponding to the merchant name to obtain the recommendation model; wherein the actual category name is the actual merchant category name corresponding to the merchant name of the user to be trained.

15. The method according to any one of claims 10 to 14, characterized in that: Before determining at least one target merchant name corresponding to the merchant name from a preset training set, the method further includes: The merchant name is preprocessed to obtain a processed merchant name.

16. A device for recommending merchant information, characterized in that: include: A determination module, used to obtain a merchant name of a user, and determine at least one target merchant name corresponding to the merchant name from a preset merchant library; wherein the preset merchant library includes at least one preset merchant name; and the target merchant name is a preset merchant name associated with the merchant name; A generating module, used to generate prompt words according to each of the target merchant names; wherein the prompt words include the merchant category name and context information of each of the target merchant names; A processing module, used for performing language processing on the prompt word to obtain the merchant information of the user; wherein the merchant information represents the category code of the merchant to be recommended; The recommendation module is used to recommend the user's business information to the user.

17. A model training device for merchant information recommendation, characterized in that: include: A determination module, used to obtain the merchant name of the user to be trained, and determine at least one target merchant name corresponding to the merchant name from a preset training set; wherein the preset merchant library includes at least one preset merchant name; the target merchant name is a preset merchant name associated with the merchant name; A generating module, used to generate prompt words according to each of the target merchant names; wherein the prompt words include the merchant category name and context information of each of the target merchant names; A training module is used to train the initial model according to the prompt word to obtain a recommendation model; wherein the recommendation model is used to process the merchant name of the user as described in claim 16 to obtain the merchant information of the user; the merchant information represents the merchant category code to be recommended.

18. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 15.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 15 when executed by a processor.

20. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 15 when being executed by a processor.

Images